Randomised controlled trial of a digitally assisted low intensity intervention to promote personal recovery in persisting psychosis: SMART-Therapy study protocol

This study (acronym SMART-Therapy) is part of a broader research program (Self-Management and Recovery Technology) developing and trialling digital resources for specialist mental health services. It is a randomised controlled trial of a discrete psychosocial intervention to promote personal recovery in people with persisting psychosis. The intervention adopts a low intensity delivery model blending a course of individual face-to-face sessions with use of a website featuring lived experience content on recovery. The primary objective is to determine if this intervention will be superior to a time-equivalent comparison intervention that incorporates exploration of non-recovery-related digital resources on a measure of consumer-defined personal recovery. We will also examine a number of secondary outcomes, interview people about their experience of the intervention, use data as part of a broader economic evaluation, and examine mechanisms of change.

In examining the mechanisms of the intervention, we hypothesise effects on personal recovery via two measurable processes: (a) mental illness self-stigma and (b) self-efficacy for mental health recovery. Self-stigma, also referred to as internalised stigma, refers to the extent to which an individual adopts negative community stereotypes about mental illness and the extent to which these exert an influence on experience and behaviour [43]. For persons with persisting psychosis, these may include stereotypes of being fundamentally different from others, of inherent dangerousness, of pessimistic long-term outcomes, and of being insufficiently capable of fulfilling social roles such as parenting and work [44]. Internalised stigma includes components such as negative stereotype endorsement, alienation and social withdrawal [45]. In studies of public stigma, contact with persons with mental illness that undermines negative stereotypes and humanises persons with a mental illness appears effective in reducing negative stereotypes [46]. We hypothesise that positive portrayal of peers may have a parallel effect on internalised stigma [41].

A concept derived from social learning theory, self-efficacy refers to the expectancies that an individual has that they will be able to successfully engage in a target behaviour [47]. Self-efficacy is a robust predictor of health behaviour change [48]. In line with social learning theory we predict that observing peers discussing engagement in recovery oriented behaviours will facilitate behaviour change in our participants via increased expectancies of being able to engage in similar behaviours.

The SMART-Therapy trial is designed as a randomised, controlled, assessor-blinded superiority trial, with two parallel groups, using a 1:1 allocation ratio. The two groups both receive eight sessions with a mental health worker in addition to their routine care during a three month window, with one group being delivered the recovery intervention, and the other being delivered a befriending intervention as a comparison condition. Outcome measures are completed at baseline (prior to randomisation) and are repeated at 3, 6 and 9 months following baseline.

The study is being conducted across multiple public community mental health services in the state of Victoria, Australia, including clinical mental health services, and not-for-profit organisation-run community mental health support services. Potential participants are identified via referral from practitioners in collaborating services, supplemented by review of case-lists, plus advertising within services, and in print and online social media.

Inclusion criteria are: (a) age between 18 and 65 years inclusive; (b) diagnosis of a nonorganic psychotic disorder (schizophrenia-related disorder or bipolar disorder or major depressive disorder with psychotic features present within the past 2 years), confirmed using the Structured Clinical Interview for DSM-IV-TR Axis I Disorders (SCID) [49]; (c) sufficient conversational English for meaningful participation; (d) overall intellectual functioning within normal limits (having an IQ greater than 70, as estimated by the Wechsler Test of Adult Reading (WTAR) [50]) to ensure they have the cognitive capacity to provide consent, in addition to being able to adequately engage with the intervention. Exclusion criteria are: (e) initiation of a new antipsychotic medication, or commencement or completion of a formal psychological treatment, within the previous 8 weeks; (f) inpatient admission within the previous 8 weeks.

The recruitment target is 148 participants, providing 80 % power to detect medium effect sizes (d = 0.5) at α = .05 with 15 % attrition. This was based on an average effect size of d = 0.53 observed in a meta-analysis of online interventions [51].

Both conditions involve delivery of eight face-to-face one-to-one sessions of 50 min duration. Interventions are delivered by mental health support workers who are seconded to the project from roles in the not-for-profit community mental health support services sector. Workers in this sector were chosen to ensure that the intervention is suitable for delivery by the whole mental health workforce, rather than being restricted to postgraduate trained health professionals. These mental health workers are required to have a certificate level of training in mental health and at least 2 years of experience. They facilitate sessions for both conditions. Sessions are held alongside the participant’s usual routine treatment and care, which continues to be provided by their existing treating team without restriction from participation in the trial. Routine care would typically include antipsychotic medication, crisis support, symptom monitoring and case management, and may include receipt of formal psychological therapies.

SMART sessions are structured around use of the SMART website on a tablet computer, which is used by the facilitator and participant together during sessions. The SMART website is Drupal-based, optimised for tablet computer or smartphone use, and accessible via any internet browser including home computer.

Content domains are primarily based on the Connectedness–Hope–Identity–Meaning–Empowerment (CHIME) framework, which has been derived as a synthesis of consumer accounts of key processes involved in recovery [35]. This is supplemented by basic symptom monitoring, coping enhancement, and behaviour change material from cognitive behavioural therapies. This creates seven self-management and recovery topics: (a) recovery, (b) managing stress, (c) physical health, (d) me, (e) empowerment, (f) relationships, and (g) life. Material for each topic is organized primarily around a series of videos, including a video introduction to the topic by a mental health consumer leader, followed by a combination of videos, text, and reflective exercises. The main videos are 2–3 min in length, each featuring a selection of persons with lived experience of psychosis reflecting on their experiences in relation to a particular aspect of recovery, and how they have navigated their mental health recovery in relation to this. Topics also include expert videos (a combination of peer leaders and non-peer academics) discussing conceptual points in more detail. Participants can share comments on each video with other users of the website in a comments feed. Due to the highly individual nature of recovery, and feedback during the consultation phase, resources were developed to include a pool of material that would be utilised in a flexible manner according to the participant’s own goals and recovery priorities. Participants’ entries for key exercises in each topic populate a “roadmap” page acting as a personalised summary of key points. Each topic also concludes by encouraging the person to identify key changes they intend to make, used to populate a task list. The site also contains a charting tool for self-monitoring stress, mood, sleep, physical health and self-esteem, and a peer-moderated member forum. Further details on the site development are detailed elsewhere [40].

Sessions involve use of the website to stimulate discussion about self-management and recovery. The worker has the following roles: (a) using website material to encourage the participant to reflect upon their own recovery, (b) facilitating the participant forming intentions to make changes on the basis of what has been discussed, (c) promoting familiarity with the site to independently use its features, and (d) encouraging the participant to use the site between sessions. The initial session includes orienting the participant to both the website and the concept of personal recovery. Participants are provided with their own login to the SMART website, and are introduced to the navigation, supported by a video on the site. This is followed by videos on the concept of personal recovery, and discussion about different material included in the website. This is used to identify priorities in the material for the therapist and participant to look at in subsequent sessions. In subsequent sessions, the worker and participant identify the area they want to focus on, with the core of the session involving a process of viewing website material together, using this as a prompt for discussion about the participant’s own recovery, and an invitation to complete reflective exercises on the site and/or make public posts about the content if willing. These sessions begin with a review of reflections since the last meeting, and any content which the person has used, and conclude by considering any changes the person would like to enact, and planning ongoing use of the site during the week.

Participants in the comparison condition receive sessions based upon a neutral social contact, involving shared activity and discussion about non-mental health topics such as current affairs, recent events and activities in the participant’s life, and the participant’s interests. This is a manualised intervention, the structure of which is based upon the befriending protocol of Bendall et al. [52], a credible and acceptable comparison condition [53], which has been used as a control condition to match for therapist time in a number of psychosocial intervention trials [54‐56]. In this trial the standard protocol is augmented by use of the tablet computer to browse Internet sites related to participants’ interests, and as a potential means of facilitating discussion. This comparison condition was selected to provide a control for both therapist time, and for engagement in the use of technology.

In both groups, interventions are continued until, (a) all eight sessions are completed, (b) the end of the three month therapy delivery window is reached, or (c) the participant chooses to withdraw. If participants experience significant distress or deterioration in their mental state, researchers remind participants of their right to withdraw, and encourage them to make an informed choice about their ongoing participation. If participants wish to discontinue intervention sessions, they are still followed up for assessment if willing to do so.

Participant adherence to protocols will be monitored through records of session attendance. In secondary analyses, attendance of at least four sessions will be considered as the threshold for receiving a minimum dose of the intervention. Workers attend weekly supervision led by NT to ensure protocol fidelity. Worker fidelity will be further assessed by a researcher blind to participant allocation listening to a random sample of audio recordings of sessions, and being asked to allocate sessions to condition.

The trial uses minimisation randomisation, conducted using the software Minimpy (http://​sourceforge.​net/​projects/​minimpy). This generates random allocations based on the stratification parameters for each participant enrolled in the study. The randomisation procedure and setup of the Minimpy parameters was prepared by the study statistician to stratify by type of mental health service the person receives (clinical vs community sector vs both), symptom severity (Positive and Negative Syndrome Scale (PANSS) total ≥ vs < 67), and frequency of Internet use (daily vs less than daily). Following commencement of the trial, the software is operated by a person independent of the research team, who receives participant codes from the research team as each baseline assessment is completed, runs the allocation, and advises of allocations via return email.

Research staff conducting assessments are blind to participant allocation throughout the study. Participants are unavoidably aware of the conditions they have been randomised to. To reduce expectancy effects for superiority of the intervention over the control condition, the trial was framed as a comparison between two ways of integrating technology into mental health service delivery, with a health condition using technology to discuss mental health recovery, and a social condition using technology to support discussion about the person’s interests. To maintain assessor blinding, participants are regularly reminded not to divulge their allocated condition or any details of their intervention to assessment staff, as are their mental health workers. Breaches in blindness are recorded and addressed by changing the rater whenever possible. Blindness is also assessed at each assessment following the intervention by asking raters to nominate a guessed treatment condition for the participant and to indicate their level of confidence. The principal investigator (NT) and project manager (FF) are not blinded, and respond to any clinical and research issues during the trial that require knowledge of a participant’s’ condition.

A schedule of assessments is provided in Table 1. Assessments are performed as close as possible to the allocated time-point. If the assessment cannot be performed within 6 weeks of the allocated time, this assessment is considered missed and the next assessment is performed at the allocated time.

Data are initially recorded on paper forms which are securely stored and extracted into password protected electronic files. Data extraction is conducted by research staff blind to participant allocation. Procedures to identify potential errors include double entry from selected paper files, range checks, and examination of outliers.

The primary focus of the analysis is differential changes in the SMART group versus Befriending over the full post intervention period. Analyses will be undertaken using mixed-model repeated measures (MMRM) allowing for autoregressive dependence. MMRM is well-suited to ITT analyses because this approach uses all available information from subjects. Participants with incomplete data are not discarded and missing data are not replaced with estimated values or observations carried forward. Instead maximum-likelihood estimation is applied with the available data. Where there are significant group x time interactions, planned contrasts will compare changes from baseline under each intervention.

Where outcome measures exhibit non-normal distributions across time points, appropriate transformations will be applied. If any of the outcome measures show significant baseline differences between SMART and Befriending, and between completers and noncompleters, corrections for sample imbalance and attrition bias will be applied using propensity scoring. This will involve firstly constructing a logistic regression model with treatment condition as the outcome, and baseline clinical variables and demographics as predictors, plus a “missing-at-post-therapy” variable to index differential effects of attrition. Secondly the effects of any imbalance or attrition will be addressed by including the estimated probabilities from this model (propensity scores) as a covariate in analyses.

The differential effects on outcomes of the process variables (self-efficacy and self-stigma) and medication changes will be tested across conditions with a maximum likelihood Hierarchical Linear Modelling analysis. In these models time will be treated as a linear covariate. Baseline measures for cognitive function, recovery style, intervention engagement, diagnosis, demographics, baseline internet use and expectancy will be used as covariates, while testing interaction effects for condition and the baseline social engagement moderators (loneliness, attachment, peer group identification and peer contact). Structural equation modelling will then be used to demonstrate the mediation mechanism of outcome improvement for each condition separately, with bootstrap analyses used to obtain 95 % confidence intervals for all effect sizes.

Data will also be used as part of a broader economic evaluation of the use of the website involving cost-consequences analyses whereby incremental costs of interventions using the site are compared with the full spectrum of outcomes measured vs relevant comparators. For this study, QALYs will be calculated from AQoL8D data, and costs will be derived from estimates of the costs involved in running the website (website development and maintenance costs, forum moderation, and therapist activity data), adjusted for any differences in resource use between conditions. Standardised economic evaluation techniques will be used including incremental analysis of mean differences, with confidence intervals derived using bootstrapping.